Chapter 6

Control of Microstructure and Properties

6.1. Introduction

The theory presented in the previous chapters will now be used to more closely examine some welded joints in an attempt to deduce the microstructures and properties that are likely to appear. This is done by looking at some specific examples and analyzing the thermal cycles and the choice of base material and consumables. The risks of cracking will also be assessed. We begin by focusing on the thermal cycle.

6.2. Predictions from the Thermal Cycle

Two typical welding processes are considered to illustrate the problem. The first represents the use of relatively high heat input welding for thin plates, while the second relates to low heat input welding in relatively thick plates. The relevant equations were presented in Section 1.7.

The joint configurations used are shown in Figure 6.1. In the first case the plate thickness is 12.5 mm, and the welding process is single-wire submerged arc welding (SAW). The configuration shown in Figure 6.1, an I-joint, is used when the mechanical properties demands are moderate. If better mechanical properties are required, different variants of Y-joints are used. In the second case gas-metal arc welding (GMAW) is used for joining a 35-mm-thick plate. Here, a V-joint is assumed. This joint type is used when the welding can only be made from one side. Otherwise, it is common to use X-joints. However, for the analysis made here, the exact choice of joint geometry is irrelevant. A first question may be how accurately the positions of the isotherms can be found, due to differences in the values of the physical constants between different materials. This will now be examined by looking at three different steels.

It is necessary to determine whether the welding conditions will lead to two-or three-dimensional cooling conditions. As pointed out in Section 1.7, the heat flow equations can be simplified considerably if it is assumed that the welding speed is so high that the heat flow in the direction of welding can be neglected. This means that the heat flow will be either one-dimensional (in thin plates, where heat is only flowing sideways) or two-dimensional (in thick plates, where the heat flow is directed both downward and sideways).

To quickly determine which heat flow mode operates, the diagram of Myhr and Grong, shown in Figure 1.39 and reproduced in Figure 6.2, is used. The dimensionless parameters are δ = vd/2κ and θ_p/n. Here, θ_p is the normalized peak temperature (T_p − T₀)/(T_m − T₀), T_p is the peak temperature, T_m is the melting temperature, T₀ is the preheat temperature, and n = (qv)/(4πκ²ρc(T_m − T₀)).

FIGURE 6.1. The joint configurations used for the two case studies. An I-joint is often used in SAW of relatively thin plates, when the demands on mechanical properties (especially toughness) are moderate. The second joint configuration is a 60° V-joint, which is relatively common when welding thicker plates.

FIGURE 6.2. Heat flow map, showing the dominant modes of heat flow. The points indicated show the calculated values for SAW and GMAW of the three different kind of steels. The dominant mode is the thick-plate solution, two-dimensional heat flow.

TABLE 6.1
Values of the Physical Constants

Alloy	K (mm2/s)	c(J/°C g)	λ (W/mm) (°C)	ρ (g/mm3)
Ordinary carbon steel	8	6.1 × 10-7	0.038	7800
C-Mn steel	4.6	7.2 × 10-7	0.026	7800
Low-alloy steel	3	8.1 × 10-7	0.019	7800

TABLE 6.2
Welding Conditions for Two Cases

Process	Joint	Voltage (V)	Current (A)	Welding speed (mm/s)	Preheat temp (°C)
SAW	I	30	600	6.8	25
GMAW (root bead)	V	24	180	4.5	150
GMAW (fill beads)	V	30	300	4.5	150

The three different steels considered are a plain carbon steel, a C-Mn steel, and one low-alloyed steel. The values of the relevant physical constants are given in Table 6.1. The welding conditions are summarized in Table 6.2.

Figure 6.2 gives the results of the calculations for the three steels and the two welding conditions, for a peak temperature of 1300°C (2350°F). As can be seen, all points fall in the area of thick-plate solution, two-dimensional heat flow. The equation used to calculate the positions of the isotherms of different peak temperatures is then the same as Equation 1.13:

rp=(2/πe)1/2 [(q′/v)/(ρc(Tp−T0))]1/2(6.1)

The calculated positions of the isotherms for the SAW in an ordinary carbon steel are shown in Figure 6.3. It is assumed that the isotherms are semicircular in shape. This is a good approximation for the isotherms corresponding to temperatures some distance away from the fusion line. However, the fusion line itself often deviates from the ideal shape near the top side of the plate. A more common shape of the fusion line is indicated in Figure 6.3a.

Figure 6.3b shows the isotherms corresponding to three temperatures (1500, 1100, and 700°C; 2700, 2000, and 1300°F) for the three different steels. As can be seen, the differences are relatively minor (1 to 2 mm), but the limitations on how accurately the isotherms can be calculated must always be noted. The exact position of the isotherms is further affected by the temperature dependence of the physical constants.

FIGURE 6.3. (a) The calculated isotherms (using Equation 6.1) for the SAW weld in an ordinary carbon steel. In this type of weld, the fusion line usually deviates from the circular shape near the top of the weld. The dotted lines indicate the undisturbed traces of the isotherms. (b) The isotherms corresponding to three temperatures (1500, 1100, and 700°C) (2700, 2000, and BOOT) for three different steels. A circular shape of the fusion line was assumed. The carbon and the C-Mn steels have fairly close isotherms, while the low-alloy steel deviates significantly from the other two.

When the approximate position of an isotherm has been determined, the thermal cycle in a specific point can be determined, using Equation 1.11

T=T0+(q′/v)/(2πλt)  ​exp(−b/4κt)(6.2)

where b = y² + z².

FIGURE 6.4. The calculated thermal cycle at r = 9 mm (corresponding to a peak temperature of 1300°C, 2350°F) for a SAW C-Mn steel (using Equation 6.2 and the welding conditions from Table 6.2).

Taking the data for a C-Mn steel and analyzing the point r = 9 mm (corresponding to a peak temperature of 1300°C, 2350°F) gives the thermal cycle shown in Figure 6.4. The cooling time from 800 to 500°C can then be calculated using Equation 6.3, which is the same as Equation 1.15:

Δt8/5=(q′/v)/(2πλ)[1/(500−T0)−1/(800−T0)](6.3)

giving Δt_8/5 as 12.5 s. The cooling time will vary from 8.6 s in an ordinary carbon steel to 17.1 s in a low-alloy steel. This of course is a relatively large variation in cooling time and can significantly alter the microstructure and properties of the heat-affected zone (HAZ).

The thick-plate solution with two-dimensional heat flow is also applicable in the second joint. In the root bead Δt_8/5 can be calculated to 6 s, while it is 12 s for the filling passes (for a C-Mn steel). The heat input rate is only about 0.7 kJ/mm in the root run and about 1.5 kJ/mm for the other runs. The thermal cycles for the root and fill runs are shown in Figure 6.5.

From Figure 6.1, the circumstances under which one-dimensional heat flow exists in steel welds can be speculated. The value of the δ-parameter should be significantly less than 10. Taking δ = 2 as a reasonable value, then, because ν often is around 4 mm/s or above, the plate thickness (in mm) will numerically be of the same order as κ (in mm2/s), i.e., between 2 and 8 mm. Because thin plates are welded with a lower heat input, the value of θ_p/n will be high, and although the value of 8 is low, it is possible that the welding conditions still will differ from the one-dimensional heat flow area.

FIGURE 6.5. The calculated thermal cycles for the root and fill runs (peak temperature 1300°C, 2350°F) for GMAW C-Mn steel. The welding parameters are given in Table 6.2.

Using these fairly simple calculations, which can be made easily on a programmable pocket calculator, a sense of the thermal cycle the material has experienced can be gained. However, it must be stressed that these values are only for guidance, because

• The shape of the weld pool is dependent upon the current and voltage balance; thus calibration against a weld made with the actual procedure should be made.

• The values of the physical constants vary with temperature and material.

6.3. Metallurgical Effects

Once the thermal cycle for a certain procedure has been determined as closely as possible, the metallurgical reactions can be assessed. To exemplify the analysis, three steels of different qualities will be chosen and welded according to two different procedures. The selection of consumables will also be discussed, in order to assess the reactions in the weld metal. The metallurgical features that will be discussed are microstructure, mechanical properties, and defects. The following three steels are selected:

1. An ordinary line pipe steel quality, according to API X-60 specification, produced by a thermomechanical controlled processing (TMCP) rolling procedure

2. A normalized, fine-grain-treated C-Mn steel

3. A titanium-treated TMCP steel

It is emphasized that these steels are not interchangable in practice. For the line pipe application a steel of the first type mentioned above is naturally selected, while the other steel types are used in other structural applications. The third steel has a composition which is typical for TMCP steels in larger thicknesses. The steels are selected to illustrate how the selection of steels and consumables can be analyzed and not how they are applied in practical constructions. The chemical composition and mechanical properties of the steels are given in Table 6.3.

The steels may not be directly comparable, because the normalized steel has a lower yield strength and perhaps should require a thicker plate in its construction. The steels are assumed to be of good quality (i.e., having low impurity content, although the contents of arsenic, tin, and antimony are not given).

6.3.1. High Heat Input Welding

6.3.1.1. Heat-Affected Zone Effects

Two welding procedures will be analyzed: one using high heat input, resulting in a slow cooling, and one with a fairly moderate heat input, resulting in a relatively rapid cooling. As shown in Section 6.2, two-dimensional heat flow is often the most common mode for heat transport. This means that the cooling time of a weld is independent of plate thickness. The first welding procedure assumed here is SAW of a 25-mm-thick plate, using three electrodes, which gives in total a heat input rate of 4.5 kJ/tnm (115 kJ/in.). Δt_8/5 will then be around 22 s.

Because the thermal cycle in the HAZ is known, the grain growth behavior can be examined. Following the outline of Section 3.2, the austenite grain size in a steel without titanium can be found from Figure 3.8. There is a certain variation from steel to steel, mainly due to the influence of the chemical composition on the activation energy. However, this is neglected. Thus for the second steel, Figure 3.8 shows that an austenite grain size of ca. 300 μm is achieved just beside the fusion line, ca. 150 μm at a peak temperature of 1300°C (2350°F) and ca. 50 μm at 1100°C (2000°F). The lower temperature of the coarse-grained region of the HAZ is often considered to be 1100°C (2000°F). For the titanium-stabilized steels, Figure 3.13 is applicable. At the fusion line the austenite grain size is ca. 100 μm, at 1300°C (2350°F) peak temperature about 50 μm, and at 1100°C (2000°F) about 30 μm. Thus, at low peak temperatures there is no significant benefit of microalloying with titanium.

TABLE 6.3
Composition (wt%) and Mechanical Properties of Steels (E_w and P_cm Calculated According to Equations 1.8 and 1.9)

The next step is to assess the likely microstructures that will appear. This is most correctly made using a continuous cooling transformation (CCT) diagram. However, such diagrams must be obtained experimentally and are seldom at hand. From a time-temperature transformation (TTT) diagram, which can be calculated using the method proposed by Bhadeshia,¹ a first check can be made, although the transformation described by these diagrams is only for isothermal conditions. TTT diagrams for the three steels are shown in Figure 6.6. It should be noted that it is only the start of the transformation that is calculated. This is almost independent of austenite grain size and thus no influence on the diagrams is seen from the differences in austenite grain size resulting from the addition of titanium. The following characteristics may be noted (in the approximation in which isothermal diagrams are used):

• The nose of the diffusional transformation curve appears at approximately 700°C for all steels.

• In steels 1 and 2 the nose appears at 4 s, while in steel 3 it appears at 1.5 s.

• The shear transformation curve is moved slightly to the left in steel 2 compared to steel 3.

• The effect of the higher carbon content of steel 2 is only to lower the M_s temperature. The position of the transformation curves is hardly affected.

FIGURE 6.6. Calculated TTT diagrams for the three steels presented in Table 6.3: (a) X-60 line pipe steel, (b) normalized steel, and (c) TMCP steel. The calculations were made using the method of Bhadeshia.¹ Steels a and b are very similar, except for the M_s temperature. Steel c, having a lower alloying content, has the C-curve for diffusional transformation more to the left, meaning that allotriomorphic ferrite more easily forms in this steel. The M_s temperature is on the same level as for steel a, due to the low carbon content in these steels.

What consequences will this have on the microstructure? With relatively slow cooling, it is likely that some allotriomorphic ferrite is formed in all the steels, but most probably in steel 3. However, the main part of the microstructure will be lath ferrite. In all three steels the laths are most likely a mixture of Widmanstatten side plates and bainite. The higher carbon content of steel 2 will lead to a higher hardness of the bainite due to an increased proportion of carbides.

The TMCP steels will have a much finer grain size in the coarse-grained region, meaning that the chance of forming allotriomorphic ferrite is higher. Even if the starting temperature of the transformation is independent of austenite grain size, the total amount naturally is affected. Smaller austenite grains means a higher density of nucleation sites for grain boundary ferrite, leading to a higher volume fraction of allotriomorphic ferrite. Not only is the allotriomorphic ferrite affected, but laths that appear will also be finer in the titanium-alloyed steels. Thus, the normalized steel will be harder due to the higher carbon content and also more brittle due to the presence of coarser laths. On the other hand, the allotriomorphic ferrite can also induce brittleness, but because it is softer, the overall effect is likely to be in favor of the titanium-stabilized low carbon content steels.

Figure 6.6.b

Figure 6.6.c

In the regions away from the fusion line the austenite grain size will be smaller in all steels, enhancing the probability of allotriomorphic ferrite formation. Because there is practically no difference in austenite grain size between the steels here, the microstructures will be fairly similar.

In summary, this means that the lower carbon content steels will have a softer HAZ than steels with a higher carbon content. There will not be any ferrite grain refining advantages associated with this higher hardness; thus toughness of the high carbon content steel is likely to be lower. Any difference in toughness between 0.1 wt% carbon and 0.06 wt% carbon is difficult to judge. There are other factors here that also must be considered, such as differences in manganese and nickel content.

The maximum hardness of the three steels can be calculated according to the BL 70 formula of Suzuki. A common specification in offshore industry is a maximum hardness of 248 HV, to avoid the risk of stress corrosion cracking. The two low-carbon steels easily fulfill this requirement for this procedure, having calculated hardnesses of 191 (TMCP) and 205 (X-60) HV. The normalized steel has a calculated hardness of 248 HV, which is exactly the limit of the specification.

6.3.1.2. The Weld Metal

The choice of weld metal is dependent on several factors:

• The requirements for mechanical properties must naturally be met. In welds for offshore use these are typically 40 J (30 ft-lb) at −40°C (−40°F).

• The productivity must be high, as in this example obtained by using several wires.

• The profile of the weld bead must give a smooth transition to the base plate, in order to avoid fatigue cracking problems during service.

To obtain the toughness requirements, a basic flux is the natural choice. However, in many cases a semibasic flux can give satisfactory properties. This may then be a better choice, because semibasic fluxes generally have better operating characteristics for high heat input welding.

The alloy content of the weld metal is determined by the strength and toughness levels required. The most common alloying concepts used for high heat input welds are either C-Mn, C-Mn-Mo, or C-Mn-Ti-B alloying. However, irrespective of the choice of alloying elements from the consumable, there is such a large dilution with the base plate that this has a very significant influence on the composition and microstructure of the weld metal. The dilution in this type of procedure is estimated to be around 70%, i.e., 70% of the weld metal is made up from base plate material and 30% from the welding consumables.

TABLE 6.4
Chemical Composition of All-Weld Metal Samples

The chemical composition of a weld metal in an all-weld metal test differs slightly from the chemical composition of the wire. This can be due to the fact that some fluxes can be alloying, e.g., in manganese, chromium, or other elements. Another factor is the creation of a gas-metal-slag equilibrium which mainly affects the carbon content. A typical all-weld metal composition is given in Table 6.4 for the three types of weld metals commonly used.

Thus, for each of the three steel types, there will be three possible compositions of the weld metal. This gives nine possible weld metal compositions, which are listed in Table 6.5, assuming a dilution of 70%.

What can be noted from Table 6.5 is the large influence of the base plate composition. The calculations made are purely arithmetic, but are in agreement with experimental results. The carbon content may be adjusted slightly by the slag-metal equilibrium, but not to such a degree that the conclusions from this exercise are violated.

The microstructure of the weld metal is determined partly by the hardenability of the alloy (mainly determined by the amounts of carbon, manganese, molybdenum, and boron) and partly by the size and type of inclusions. The toughness in turn is determined by the general microstructure, the size of the inclusion, and also the impurity element content.

A passing comment can be made about a common misunderstanding of the homogeneity of weld metal composition. It is sometimes believed that there is a gradient in chemical composition in the weld metal, due to dilution effects, so that near the fusion line there is a concentration of elements from the base plate and near the center of the weld there is a concentration of elements from the consumables. However, the mixing of elements in the weld pool is very effective due to fluid streaming in the pool, so that within a single bead no such inhomogeneities can be found. Any inhomogeneities are mainly due to segregation effects, which exist on very local levels (between solidification dendrites). One type of gradient that may occur is of course between different beads. The beads lying close to the base metal will be richer in alloying elements diluted from the base plate.

Starting with the type of particles, Mills et al.² found that the presence of galaxite MnOAl₂O₃ is highly beneficial. Although the reaction sequence in the weld pool for oxide inclusion formation is not known in detail, a simple model can be used, in which it is assumed that aluminum first reacts with oxygen, then titanium, and then manganese, as outlined in Section 4.3.

TABLE 6.5
Chemical Composition of Weld Metal Sample, Assuming a Dilution of 70%

Assuming that the soluble aluminum content is 0.006 wt%, then Table 6.6 can be created. Here the amount of oxygen tied up in alumina is calculated, assuming that the oxygen content of the A1₂O₃ is given by 1.13 times the aluminum content. The remaining oxygen is then found by subtracting the amount tied up in alumina from the total oxygen.

With the highest aluminum content (0.025 wt%) there will be an excess of free aluminum. This has two negative consequences. First, only Al₂O₃ will form and this particle is considered to be a poor nucleant for acicular ferrite. Second, the free excess aluminum is suspected to be detrimental to the impact toughness properties.³ The optimum aluminum content is around 0.010 wt% if the oxygen content is around 150 ppm. Then there is enough oxygen to form both Al₂O₃ and MnO, leading to a tendency to form galaxite. This, however, also requires that the titanium content is low; otherwise, titanium oxides are formed in preference to MnO and thus no galaxite is formed.

If the titanium content is high, as when a Ti-B consumable is used, then titanium oxides probably are the nucleants for acicular ferrite, instead of galaxite. The aluminum content must then be kept low (<0.005 wt%) in order to ensure maximum formation of titanium oxides. It is still unclear which titanium oxide type is actually the most effective ferrite nucleant and therefore the optimum oxygen to titanium ratio cannot be stated. It must also be considered that some titanium is tied up with nitrogen (Equation 4.4). However, TiN is also considered to be an effective nucleant, so from this viewpoint, it may not present any drawback if some of the titanium is used for nitride formation. This also has the positive effect of removing some nitrogen from solid solution, which should improve the toughness. Thus, to ensure that the amount of acicular ferrite is maximized, it is important that the content of the microalloying elements both in the plate and in the consumable is adjusted so that particles of the correct kind can be formed and that the ratio between elements such as aluminum, titanium, oxygen, and nitrogen is kept close to the optimum value. Knowledge in this area is still rather limited and much more research is clearly needed.

TABLE 6.6
Oxygen Content Tied Up in Aluminum Inclusions ([O]_i,A1)

The hardenability of the alloys must be adjusted to ensure that high amounts of acicular ferrite are formed. However, carbon content in excess of about 0.10 wt% is generally avoided. The microstructure of the nine weld metals from Table 6.5 has been calculated with the model of Bhadeshia et al.⁴ for the welding conditions specified above and the result is shown in Figure 6.7. The large effect of the carbon content can easily be seen. However, 0.14 wt% carbon is an unusual and unrealistic value. There is no effect of molybdenum on the microstructure, which is somewhat surprising, because this element is frequently used in this kind of application. However, with the high manganese contents present here, the hardenability is still so high that the small amount of molybdenum does not influence the microstructure.

The most effective way to maximize the acicular ferrite content is by using titanium and boron. Boron prevents the formation of allotriomorphic ferrite, enhancing the possibility for formation of acicular ferrite. However, it has been found that addition of titanium is necessary. Without titanium, bainite easily forms instead of acicular ferrite. In the alloys above, the boron content is too low to achieve maximum effect. This is due to the high dilution and it is customary to have special wires that give about 100 ppm of boron in an undiluted weld metal, resulting in about 30 ppm of boron in the diluted weld metal.

The tensile properties of the weld metal can be calculated by the method of Svensson et al.⁵ The values are shown in Table 6.7. In this case it is assumed that the testing is made in the last deposited bead. This means that only the primary structure of the weld metal will contribute to strength and not the reheated zones.

FIGURE 6.7. The microstructure of the nine weld metals from Table 6.5 as calculated using the model of Bhadeshia et al.⁴

The impact properties can be assumed to be proportional to the amount of acicular ferrite, because the size of the particles is probably similar in all the weld metals. The size of the particles is mainly determined by the heat input. The arithmetic mean three-dimensional diameter can be estimated from Figure 4.15 to be about 0.6 μm.

Finally, the risk for solidification cracking must be discussed. In modern steels such as the ones used in this example, which have low impurity content, this risk is very low. However, under certain circumstances, as when the depth-to-width ratio is larger than one, the risk may be large even for these steels. The UCS parameter, calculated according to Equation 1.7, gives the values shown in Table 6.8.

The highest risk for solidification cracking exists in the weld metal in the normalized steel, mainly due to the high carbon content. No single weld metal has a UCS value above 30, where the solidification cracking risk is very large. However, values around 20 also call for a careful inspection of the welding procedure to ensure that no circumstances enforcing cracking are present. Normally, when making these kinds of welds, with one run from either side, the welds are relatively deep to ensure penetration, which increases the solidification cracking risk.

TABLE 6.7
Tensile Properties of Weld Metals, Calculated by the Method of Svensson et al.⁵

Note: It is assumed that the tensile specimens are extracted from the last deposited bead and that no reheated zones will influence the results.

6.3.1.3. Further Considerations

Increased productivity is usually associated with increased heat input. The mechanical properties, and most notably toughness of the welded joint, are influenced significantly by this variation in heat input. Moreover, different steels respond differently to heat input variations due to differences in the transformation characteristics. This means that a variety of transformation products appear when the cooling time is varied.

A number of investigations that discuss the toughness of some different steels and weld metals as a function of heat input will be presented in this section. Dolby⁶ reported a major investigation on normalized steels. The toughness was reported as crack-tip opening displacement (CTOD) values of 25-mm-thick plates. Figure 6.8 shows the toughness at -60°C (-76°F) in the coarse-grained HAZ for two different steel types, one plain C-Mn steel with 0.17C-l.lMn (wt%) and one 0.15C-1.4Mn-Nb steel. The niobium content was fairly high, about 0.06 wt%. As can be seen, the plain C-Mn steel had a low toughness for low heat inputs (due to martensite formation), but for heat inputs between 3 and 7 kJ/mm (75 and 175 kJ/in.), the toughness was good. An extremely high heat input (25 kJ/mm, 635 kJ/in.) led to a decreasing toughness. In the other steel, the toughness was low for low heat inputs, just as for the C-Mn steel. However, increasing heat input led to an increase in toughness, but only up to 2 kJ/mm (50 kJ/in.). Higher heat inputs led to a rapid decrease in toughness.

TABLE 6.8
Value of the UCS Parameter for Nine Weld Metals

FIGURE 6.8. Toughness at -60°C in the coarse-grained HAZ for two different steel types, one with 0.17C-l.lMn and one with 0.15C-1.4Mn-Nb. (From Dolby, R. E., Weld. Res. Suppl., 58 (August), 225-s, 1979. With permission.)

In the C-Mn steel the microstructure consisted of side plates at high heat inputs. However, these were fairly soft (the hardness was below 200 HV10), explaining the good toughness. In the C-Mn-Nb steel, niobium was assumed to promote an upper bainite-type microstructure with a somewhat higher hardness (20 to 40 HV10 harder than a nonniobium steel) and this was expected to yield lower toughness. Precipitation of niobium carbides and nitrides is commonly expected, but according to Dolby it was not possible to verify precipitation.

TABLE 6.9
Chemical Composition (wt%) and Mechanical Properties of TMCP Steel Used in the Investigation by Nevasmaa et al.⁷

In a recent investigation, Nevasmaa et al.⁷ looked at the impact toughness of the HAZ in a TMCP accelerated cooled steel. The chemical composition and mechanical properties are given in Table 6.9. The steel, 35 mm thick, was welded both with flux-cored wires and SAW. The heat input varied from 1 to 5 kJ/mm (25 to 125 kJ/in.). The result is shown in Figure 6.9. As can be seen, there is no straight relationship between heat input and toughness. However, it can be noted that, except for a few odd values, the impact toughness was above the 40-J (30-ft-lb) level, irrespective of heat input. It can further be seen that the impact toughness average values are somewhat lower for low heat inputs (1 and 2 kJ/mm, 25 and 50 kJ/in.) (although there are some exceptions) than for high heat inputs. On the other hand, the scatter in impact toughness is much higher for high heat inputs. There are two points that deviate from the general pattern. These were welded with a high heat input, but still show the features typical of low heat input (i.e., somewhat lower toughness and small scatter). In these specimens, the addition of metal powder was used to increase productivity. To obtain the same productivity without addition of metal powder, the heat input must be approximately doubled. The melting of the metal powder requires some energy, and thus the welding is effectively carried out with a lower heat input than that calculated from the welding parameters. The good impact toughness found was related to a fine-grained ferritic microstructure, partly brought about by a fine austenite grain size, due to particle-inhibited grain growth (Figure 6.10).

In the two investigations reported above, there is only one case where a drawback to high heat inputs was found with regard to HAZ toughness: when steels with high microalloying contents are used. This is true at least when the heat input is below 7 kJ/mm (175 kJ/in.).

In an investigation similar to that on the accelerated cooled TMCP steel, a direct quench and tempered (DQT) steel of higher strength (about 690 MPa, 100,000 psi) was welded with different heat inputs.⁸ The plate thickness was 40 mm and the chemical composition is given in Table 6.10. The impact toughness results are shown in Figure 6.11. The impact toughness was in general on a lower level than for the lower strength steel, which is not surprising. However, there is no trend at all in the relation between toughness and heat input, although it seems to be slightly better to use a low heat input.

FIGURE 6.9. Toughness in the HAZ of the TMCP accelerated cooled steel as a function of heat input. (From Nevasmaa, P. et al., VTT Res. Notes 1410, VTT, Espoo, Finland, 1992. With permission.)

Changes in productivity also affect the weld metal properties. For the cases reported above, the requirements on toughness can be met by using C-Mn-Ni-alloyed weld metals for the lower strength level, while a C-Mn-Ni-Mo-Cr-alloyed weld metal is needed to match the strength of the 690 MPa strength level. Both of these weld metals also match the toughness properties of the HAZ.

Another example where productivity requirements can affect the properties is when changing from single-wire SAW to welding with parallel wire. Figure 6.12 shows the relation between impact toughness at −60°C (−76°F) with heat input for four different welding procedures. As can be seen in Figure 6.13a to d, there is a difference in the amount of refined microstructure between the tests. With increasing heat input in the parallel wire procedures the amount of primary, unrefined microstructure increases. This is probably the most likely explanation for the variation in toughness. It should however be noted that in all procedures a very high toughness was found. The increased productivity of the parallel wire method can certainly be used with confidence.

FIGURE 6.10. A fine-grained ferritic microstructure, brought about partly by a fine austenite grain size, due to particle-inhibited grain growth. (From Nevasmaa, P. et al., VTT Res. Notes 1410, VTT, Espoo, Finland, 1992. With permission.)

TABLE 6.10
Chemical Composition (wt%) and Mechanical Properties of DQT Steel Used by Nevasmaa et al.⁸

6.3.2. Multipass Welding

6.3.2.1. Heat-Affected Zone Effects

If the same three steels used for the example above in high heat input welding are subjected to a different welding procedure (multipass welding using GMAW, similar to that used in Section 6.2), the cooling times will be much shorter, especially for the root runs. In Section 6.2, the Δt_8/5 was found to be 6 s for the root run if a preheat temperature of 150°C (300°F) was used. From BS 5135:1984, it can be found that no preheat is necessary for either plate, provided the hydrogen content is below 5 ml/100 g weld metal. If the hydrogen content is between 5 and 10 ml/100 g weld metal, then 100°C (200°F) preheat is necessary for the root run in the normalized steel. When a heat input rate of 1.5 kJ/mm (38 kJ/in.) is used, no preheat is necessary. Thus, to determine the cooling time for this actual case, T₀ is set to 25°C (80°F) in general usage and to 100°C (200°F) for the root run in the normalized steel. The result is given in Table 6.11. In practice, the cooling times for the fill runs will be longer than those given in Table 6.11, because the interpass temperature will be higher than 25°C (80°F). Short cooling times may be found if the welding is temporarily stopped (e.g., overnight) and then restarted without any preheating.

FIGURE 6.11. Impact toughness in the HAZ of a DQT steel, as a function of heat input. (From Nevasmaa, P. et al., VTT Res. Notes 1406, VTT, Espoo, Finland, 1992. With permission.)

FIGURE 6.12. The relation between impact toughness at –60°C (–76°F) and heat input for four different welding procedures.

FIGURE 6.13. The difference in the amount of refined microstructure between four welding procedures: (a) single wire; (b) parallel wire, 1.1 kJ/mm heat input; (c) parallel wire, 1.4 kj/mm heat input; (d) parallel wire, 2.1 kJ/mm heat input.

TABLE 6.11
Cooling Times from 800 to 500°C for the GMAW Procedure, Specified in Table 6.2, but with 100°C (200°F) Initial Temperature for the Root Run in the Normalized Steel and 25°C (80°F) Otherwise

Steel	Root run	Fill run
X-60 and TMCP	3.6	7.5
Normalized	4.7	7.5

The austenite grain growth behavior is examined using Figures 3.8 and 3.13. A grain size of 50 μm is found in the titanium steels beside the root beads at the fusion line and about 60 μm for the other passes. For the non-titanium steel the austenite grain size is about 100 μm in the root passes and about 200 μm for the filling passes. The effect of titanium on the austenite grain size is much larger for this welding procedure than for the high heat input weld. This is because during high heat input welding, the titanium particles coarsen and are less effective in restricting grain growth.

There is certainly a risk of martensite formation in the first two steels, especially near the root runs, due to the combination of short cooling times and the relatively long time taken for transformation to start. However, as seen from the TTT diagrams (Figure 6.6), the M_s temperature is higher in the lower carbon concentration steels. Any martensite formed then has a good chance of being autotempered during further cooling. Thus, near the root beads it is possible that there will be a martensitic microstructure, which is either untempered (for the normalized steel) or tempered (for the lower carbon concentration steels). For the filling passes, the cooling time is just above that needed to avoid martensite formation and no preheat is necessary. The microstructure will be of the lath type, mainly bainitic. The finer austenite grains in the titanium-alloyed steels give a finer packet size for the bainite.

The mechanical properties of the HAZ for these plate thicknesses are commonly measured both in the root and in the top. In the root region, the steels with autotempered martensite will have the best toughness. In the top region, the titanium-alloyed steel will probably have the best toughness due to the finer bainitic structure. The normalized steel will probably have the worst properties, because, first, the bainite will be quite coarse and, second, it will be hard due to the high carbon content. Toughness is, as noted previously, the most difficult property to assess. General conclusions such as those above can be drawn, but it is not possible to put more quantitative figures on the toughness. Data on toughness for procedures as above always show large scatter. However, most steels used in offshore structures, as well as the normalized steels used during the 1970s, show adequate toughness if the welding procedures were correctly designed. Occasional values of low toughness were connected to local brittle zones (LBZs) in the HAZ, a problem that still attracts interest. In the above procedures, the presence of LBZs has not been discussed, but toughness has mainly been assessed from the microstructure arising from a single pass. It must be strongly emphasized that, especially in the second welding procedure, the presence of many beads can significantly affect the toughness. If no LBZs appear, then multipass welding will generally improve the toughness.

The maximum hardness can be estimated with Suzuki’s BL 70 formula. Using this (Table 6.12) it can be concluded that the requirement on 248 HV maximum hardness is very difficult to satisfy. To be able to meet this requirement, the welding procedure must be redesigned.

6.3.2.2. The Weld Metal

The microstructure and properties of the weld metal depend to some extent on the shielding gas used. Flux-cored wires having a rutile slag system are more sensitive to the choice of shielding gas than are wires with a basic slag system. For rutile-type wires, the type of shielding gas controls both the content of alloying elements (through the loss of elements) and the oxygen content, with CO₂ gas giving a higher oxygen content and larger losses of manganese and silicon. Metal-cored wires, having only a small amount of slag-forming elements in the filling, behave approximately as rutile-type wires.

To meet the strength requirement of the steels, a wire giving a weld metal yield strength of about 500 MPa should be used, because overmatching weld metal strength is required. This is commonly achieved with wires having a typical chemical composition of 0.07C-1.3Mn-0.6Si (wt%), welded with Ar/20%CO₂ gas. In basic wires, only manganese and silicon are used as deoxidizing elements, while in rutile wires there is about 0.015 wt% titanium. The aluminum level is as low as possible, about 0.002 wt%, coming mainly from the metal strip. In rutile wires alloyed with titanium and boron, the titanium concentration is around 0.04 wt%.

TABLE 6.12
Maximum Hardness Values Calculated According to Suzuki’s BL 70 Formula

Steel	Root	Top
X-60	302	258
Normalized	384	341
TMCP	298	244

The composition of the weld metals in the welded joints will differ somewhat from the all-weld metal composition, due to the dilution with the base plate. In the root area, where the dilution is largest, higher contents of alloying elements are found. Because the content of the alloying elements manganese and silicon is quite similar between the plate and the weld metal, the dilution effect of these elements can be neglected. It is mainly the increased amounts of carbon that may be of concern, as this significantly influences the micro-structure and hardness of the root beads. Also, the dilution of elements such as aluminum, titanium, vanadium, and niobium should be monitored.

The dilution in these types of welds is typically 25 to 50%, depending on the welding parameters. Because the welding is made semiautomatically, the welder’s technique can influence the dilution. Assuming that the dilution is 40%, the root beads will have the chemical composition given in Table 6.13. As can be seen, the dilution from the base plate does not change the composition very much. The increased carbon content of the weld in the normalized steel will decrease the M_s temperature, but not affect the position of the noses in the CCT diagram. The oxygen content from a basic wire is around 450 ppm in an all-weld metal test. The aluminum diluted from the base plate will combine with about 100 ppm of oxygen. Although the oxygen content will also be diluted (to around 200 ppm), there will still be free oxygen to combine to either MnO, which can form galaxite, or to some titanium oxide. Thus, the condition for acicular ferrite formation seems fulfilled.

In the filling beads, away from the fusion boundary, the chemical composition of the weld metal is almost exactly the same as for the pure all-weld metal. The nucleation of acicular ferrite then must occur on manganese silicate particles, although these are generally believed to be less effective nuclei. It is possible that there are titanium oxides or nitrides formed as shells on these particles, aiding acicular ferrite formation, in spite of the low concentration of titanium.

The microstructure of all-weld metal basic-type wires contains appreciable amounts of acicular ferrite. It is difficult to directly compare with microstructures in weld metals from other processes, such as shielded metal arc welding (SMAW) or SAW, because the welding parameters may differ and thus the thermal conditions may vary. Even if the welding parameters were very similar, the cooling conditions would still be different, because each process has its own thermal characteristics. The toughness of weld metals from flux-cored wires is determined mainly from the oxygen content. Basic-type wires, with about 450 ppm oxygen, typically have a toughness of 50 J (37 ft-lb) at −50°C (−58°F), when alloyed with 1 wt% nickel. This is somewhat lower than a weld metal from a similarly alloyed stick electrode. The difference mainly lies in the amount of oxygen, which is about 100 ppm lower in the stick electrode weld metal.

TABLE 6.13
Chemical Composition of the Root Beads

Basic wires have relatively poorer operating characteristics in positional welding than rutile wires. The rutile weld metals on the other hand have a higher oxygen content and thus lower toughness. To improve toughness, some newly developed rutile wires are alloyed with titanium and boron. The Ti-B alloying concept is common in Japanese consumables, in stick electrodes, cored wires, and in submerged arc fluxes. In other countries, this concept is rarer and is mainly used in cored wires and occasionally in SAW wires. The microstructure of such weld metals is shown in Figure 6.14, which illustrates the high proportion of acicular ferite that is found in these weld metals. The toughness is also impressively high, about 100 J (75 ft-lb) at –40°C (–40°F), despite an oxygen content of about 600 ppm.

In these weld metals the nucleation of acicular ferrite relies on titanium oxides as nucleants. The aluminum content is kept low (below 0.002 wt%) and the nitrogen content is also very low (around 25 ppm). The chemical composition of the particles has not been investigated, but assuming that Ti₂O₃ is the effective particle and that the particles are stoichometric, 0.040 wt% titanium ties up about 400 ppm oxygen. Thus, the full effect is taken from the titanium content, but there is still a surplus of oxygen.

These wires have the disadvantage that the weld metals are sensitive to heat treatments, in that the toughness is lowered. They are therefore not recommended to be used in situations requiring stress-relieving heat treatment or hot-forming operations. Addition of nickel is frequently used to improve the toughness. Figure 6.15a shows the effect on the acicular ferrite content from the addition of nickel to a base composition of the weld metal of (wt%) 0.07C-1.3Mn-0.6Si. Figure 6.15b shows the effect of nickel on the yield strength. As can be seen, increasing nickel has a great effect on the acicular ferrite content. The toughness improvement obtained with addition of nickel thus has two sources: a refinement of the grain size and lowering of the stacking fault energy.

FIGURE 6.14. The microstructure of Ti-B-alloyed weld metals, consisting almost entirely of acicular ferrite.

GMAW with cored wires is relatively common in the U.S. and Japan. In Europe, the use of the process is rapidly increasing. Much development is also taking place, resulting in new, better consumables. Another fascinating development in this process is the ability to closely control the drop detachment with the new synergic power sources. This will increase the possibilities of positional welding with basic-type wires, enabling improved weld metal properties.

6.4. Controlled Bead Placement

Based on the original ideas of Alberry and Jones,⁹ Reed¹⁰ developed a model to analyze the possibility of achieving complete reaustenization of underlying beads. As noted previously, reaustenitized structures usually have better mechanical properties than the as-deposited microstructure. The model was developed for SMAW. To predict the positions of the isotherms more precisely than is possible with the Rosenthal equations, Reed used the extension of the Rosenthal equations made by Ashby and Easterling.¹¹ They modified the equations by letting the power intensity of the heat source be distributed over a finite area and by deducting the latent heat of fusion from the arc power.

The size and shape of a bead-on-plate weld was calculated from these modified heat flow equations and from a knowledge of the volume of deposited metal per unit length of weld. This volume is calculated from the electrode diameter, the efficiency, the feed speed of the electrode, and the welding speed by the relation

FIGURE 6.15. (a) The calculated changes in microstructure in flux-cored wire weld metals, from the alloying of nickel to a base composition of the weld metal of 0.07 C-1.3 Mn-0.6 Si. (b) The calculated effect of nickel on the yield strength of weld metals from flux-cored wires.

A1=[(π electrode diameter2)/4]× efficiency×(feed speed)/(welding speed)(6.4)

To determine the influence of the welding process on the microstructure of the HAZ (either in the weld metal or in the base plate), the following temperatures were used to describe the microstructure:

Ts	solidus temperature
Tgc	γ-grain-coarsening temperature
Ac3	temperature where the α/γ transformation is completed on heating
Ac1	temperature at which the α/γ begins on heating
Tt	tempering temperature

In a computer program, all these temperatures could be given different values, depending on steel type. The following parameters were analyzed by Reed:

• Electrode diameter

• Interpass temperature

• Arc current

• Feed speed

• Transformation temperature

It was previously noted by Clark¹² that the feed speed varied linearly with the arc current, as

feed speed=(4KI)/(πρd2)(6.5)

where K is a numerical constant (= 2.5 mg/As), I is the welding current, d is the diameter of the electrode, and ρ is the density of the electrode.

Plate 6^* shows the effect of increasing the electrode diameter d from 3.25 mm to 6 mm, keeping the current constant. As expected, more beads are needed to fill the joint with a smaller rather than a larger diameter electrode. However, what is more interesting is what happens to the underlying material. With the model, the volume fraction of unreaustenitized (columnar) structure and coarse-grained structure can be calculated. Increasing the electrode diameter leads to a slight decrease in the volume fraction of the columnar structure, while the volume fraction of the coarse-grained zone remains virtually constant. In practice, higher currents are used with larger diameter electrodes, depositing larger beads, so that even fewer beads are needed to fill the joint.

The effect of welding current, keeping all other variables constant, is shown in Plate 7.^* Fewer weld beads are needed with higher currents, and the temperature field changes below the beads. This leads to an increase in the volume fraction of columnar structure and a decrease in the volume fraction of the coarse-grained zone. An increase in the interpass temperature moves the isotherms away from the weld bead, leading to a larger degree of reaustenitization, as shown in Plate 8.^*

The computer model can now be used to find the optimum combination of variables in the welding procedure, so that as complete reaustenitization as possible is obtained. To obtain as high a proportion of reaustenitized material as possible, a low Ac₃ temperature is beneficial, as shown in Plate 9.^* Depositing very thin beads by having a very low electrode feed speed (Plate 10)^* also gives large amounts of reaustenitized material. However, it must be realized that this procedure will result in lower productivity if the welding cannot be done at very high speeds.

To summarize, it is possible to analyze with this method the influence of various variables on the weld metal and HAZ structure in multipass welds. However, these ideas are still in the beginning of their development and there is much additional research and correlations with actual welds that must be made to be able to fully utilize the method.

References

1. Bhadeshia, H. K. D. H., Thermodynamic analysis of isothermal transformation diagrams, Met. Sci., 16, 159, 1982.

2. Mills, A. R., Thewlis, G., and Whiteman, J. A., Nature of inclusions in steel weld metals and their influence on formation of acicular ferrite, Mater. Sci. Technol., 3(December), 1051, 1987.

3. Terashima, H. and Hart, P. H. M., Effect of aluminum on C-Mn steel submerged arc weld metal properties, Weld. J., 63(June), 173-s, 1984.

4. Bhadeshia, H. K. D. H., Svensson, L.-E., and Gretoft, B., A model for the development of microstructure in low-alloy steel (Fe-Mn-Si-C) weld deposits, Acta Metall., 33, 1271, 1985.

5. Svensson, L.-E., Gretoft, B., Sugden, A. A. B., and Bhadeshia, H. K. D. H., Computer-aided design of electrodes for arc welding processes. II., in Proc. Int. Conf. Computer Technology in Welding II, The Welding Institute, Abington, U.K., 1988, paper 24.

6. Dolby, R. E., HAZ toughness of structural and pressure vessel steels — improvement and prediction, Weld. J., 58(August), 225-s, 1979.

7. Nevasmaa, P., Cederberg, M., and Vilpas, M., Weldability of Accelerated-Cooled (AcC) High Strength TMCP Steel HT50, VTT Res. Notes 1410, VTT, Espoo, Finland, 1992.

8. Nevasmaa, P., Cederberg, M., and Vilpas, M., Weldability of Direct-Quenched and Tempered (DQT) High Strength Steel HT80, VTT Res. Notes 1406, VTT, Espoo, Finland, 1992.

9. Alberry, P. J.. and Jones, W. K. C., Computer model for prediction of heat-affected zone microstructures in multipass weldments, Met. Technol., 9, 419, 1982.

10. Reed, R. C., The Characterisation and Modelling of Multipass Steel Weld Heat-Affected Zones, Ph.D. thesis, University of Cambridge, Cambridge, U.K., 1990, chap. 3.

11. Ashby, M. F.. and Easterling, K. E., The transformation hardening of steel surfaces by laser beams. I. Hypo-eutectoid steels, Acta Metall, 32, 1935, 1984.

12. Clark, J. N., Manual metal arc weld modelling. I. The effect of process parameters on dimensions of weld beads, Mater. Sci. Technol., 1, 1069, 1985.

---

* Plates 6 and 7 appear following page 84.

* Plates 8 to 10 appear following page 84.